convert-from-nuget indirect refs should not override direct deps

[bartelink]

convert-from-nuget got the conversion of hyprlinkr almost perfect:

There were 2 unit tests using Moq, both specifying 4.2.1312.1622.

The converted file instead had latest (4.2.1409.1722)

This is more than likely due to, as the lock file shows, AutoFixture has an indirect reference to Moq too:

    AutoFixture.AutoMoq (3.16.4)
      AutoFixture (>= 3.16.4)
      Moq (>= 3.1)

The tests have assemblyRedirect elements pointing at the specific Moq version and test running fails when the when the version is inferred incorrectly.

One final thing is that the global packages.config didnt declare NuGet.CommandLine, so deps needs:

nuget NuGet.CommandLine

Final deps file with 2 adjustments above:

source http://nuget.org/api/v2

nuget Moq 4.2.1312.1622
nuget NuGet.CommandLine

nuget xunit.runners 1.9.2
nuget Zero29 0.5.0
nuget Microsoft.AspNet.Mvc 4.0.20710.0
nuget Microsoft.AspNet.Razor 2.0.20710.0
nuget Microsoft.AspNet.WebApi 4.0.20710.0
nuget Microsoft.AspNet.WebApi.Client 4.0.20710.0
nuget Microsoft.AspNet.WebApi.Core 4.0.20710.0
nuget Microsoft.AspNet.WebApi.WebHost 4.0.20710.0
nuget Microsoft.AspNet.WebPages 2.0.20710.0
nuget Microsoft.Net.Http 2.0.20710.0
nuget Microsoft.Web.Infrastructure 1.0.0.0
nuget Newtonsoft.Json 5.0.8
nuget Albedo 0.5.0
nuget AutoFixture 3.16.4
nuget AutoFixture.AutoMoq 3.16.4
nuget AutoFixture.Idioms 3.16.4
nuget AutoFixture.Xunit 3.16.4
nuget netfx-Guard 1.3.2.0
nuget netfx-Reflector 1.0.0.10
nuget xunit 1.9.2
nuget xunit.extensions 1.9.2

[theimowski]

what was the warning message? convert-from-nuget should take latest version from all defined in packages.config files, so maybe it's likely that one of your project had the dep on that version?

[bartelink]

No, 2 had exactly 4.2.1312.1622

But an indirect dep (Moq from AutoFixture), specifies >= 3.1

This resulted in the deps file getting 4.2.1409.1722 instead of 4.2.1312.1622

UPDATE: Can review the messages but pretty confident it is wrong (the tests fail :) )

[bartelink]

@theimowski One thing I didnt mention is that it #158 is also useful in the context of converting hyprlinkr

[theimowski]

what's the status of this? Is the --interactive mode for simplify enough to deal with the problem?

[bartelink]

I don't know - for me there is a clear algorithmically derivable solution, and I'd prefer that. But OTOH I'll be honest and say that I haven't used --interactive (I want it to do its best, warn about stuff it can warn about, giving suggestions. I can edit the file myself).

The bottom line is that the assembly redirects in the project want it to be exactly that version and an indirect dependency I have no control over caused it to agressively override 2 identical clear instructions)

But I'm not familiar with the resolution code - if this conflicts with a lot of what it does this might be very messy. And there are definitely bigger fish to fry - as #154 shows, there's plenty scope for more work on the coding rules. Fixing this is not exactly closing the final issue in resolution 4eva :)

I'd moan a lot more if other issues are closed than I would for this one, but unless someone can explain why I'm wrong I do consider it technically a bug.